[MPICH] MPICH2 ports and firewall issues - limit port range ?

[Martin Schwinzerl]

I have some problems regarding the usage of MPICH2 and would be very 
grateful, if someone here
could help me. Although I've gathered some experience with MPICH2 while 
using it at my local network,
I've only a limited understanding of ther internal workings of MPI CH2 
(and English isn't my mother tongue either),
so please bear with me :-)

I try to setup MPICH2 (ver. 1.0.3)  in an existing network consisting of 
several Linux machines (OpenSuSE 10.0)
using the Intel Compilers (C/C++ and Fortran 9.0). While the 
installation process itself turned out
to be straighforward as usual, I ran into problems bringing up a ring of 
computers using the first
mpdboot example in the install manual.

(mpdboot -n 4 -f mpd.hosts
 mpdtrace
 mpdallexit
)

After some troubleshooting (several attempts to connect computers to 
eachother using the mpdcheck program
and temporary disableing the firewalls of the computers in question), 
the local firewalls turned out to be the problem,
since they block all incoming traffic on all ports exept ssh and http.

Since providing all port information at startup of the mpd's (and 
enabling the specific ports in the firewall rules)
is definitly not suitable for our needs, and enabling all ports in bulk 
(even if, for example, they are restricted to the needed applications)
for traffic between the machines in our local network is
according to our system administrator a big no-no, I wanted to ask if 
there is some way to limit the port range
which MPICH uses , so that we can at least retreat to enable a distinct 
block of ports for our cluster. I've so far
tried to use the MPICH_PORT_RANGE environment variable to accomplish this

(e.g. mas at thph037214:~> export MPICH_PORT_RANGE="8000..9000")

but at least mpdcheck wasn't able to limit itself to ports in this range

mas at thph037214:~> mpdcheck -s
server listening at INADDR_ANY on: thph037214 1024

Any information / suggestions / hints how to limit the mpich2 programs 
to certain range of ports
(or any corrections for things I'm currently doing wrong) would be very 
welcome and appreciated!

Thank you very much for your time & efforts ,

  regards

  Martin Schwinzerl

----

Martin Schwinzerl
Institute of Physics
University Graz
Austria

http://physik.uni-graz.at/~mas

"Tomorrow is going to be wonderful, because tonight, I do not understand 
anything"
Niels Bohr